As the 6G era approaches, Industrial Internet of Things (IIoT) networks face escalating challenges in delivering real-time, energy-efficient, and low-latency services. To address these requirements, next-generation architectures are increasingly incorporating key enabling technologies such as Mobile Edge Computing (MEC), Unmanned Aerial Vehicles (UAVs), High-Altitude Platform Stations (HAPS), Digital Twins (DT), Clustered Non-Orthogonal Multiple Access (C-NOMA), and Hybrid Beamforming (HBF). Despite these advancements, significant research gaps remain in dynamic resource allocation, energy-aware task offloading, and spectral efficiency, particularly under constrained or congested terrestrial infrastructures.
Firstly, we propose a Dynamic Digital Twin Edge Air-Ground Network (D2TEAGN) architecture that leverages DT-assisted state estimation to enhance adaptive resource utilization in dynamic IIoT air-ground networks. This framework optimizes UAV trajectories, IoT device association, and task offloading, ensuring energy efficiency and low-latency service provisioning. The task offloading optimization problem is formulated as a mixed-integer nonlinear programming (MINLP) model and solved using a deep deterministic policy gradient (DDPG) algorithm, which balances computational demands and power constraints.
Secondly, we extend DT-assisted task offloading beyond UAVs by integrating HAPS as an additional service provider, enhancing network coverage and computational capacity. Unlike traditional approaches, the proposed multi-tier offloading framework allows IoT devices to offload tasks to both UAVs and HAPS, ensuring queue-aware task scheduling and efficient MEC resource utilization. This enables adaptive workload balancing, where DDPG-based reinforcement learning dynamically optimizes task scheduling and offloading based on network congestion, computational capacity, and latency constraints.
Building upon this, we further enhance scalability and dynamic load distribution by enabling UAVs to relay tasks to HAPS when overloaded, rather than solely relying on direct offloading from IoT devices. This hierarchical task management is framed as a multi-agent Markov decision process (MDP) and solved using multi-agent proximal policy optimization (MAPPO), leveraging multi-agent deep reinforcement learning (MADRL) to coordinate task offloading, minimize energy consumption, and ensure low-latency service provisioning. By introducing collaborative task scheduling across UAVs and HAPS, the framework enhances service continuity and resource optimization in dynamic IIoT applications.
Lastly, we extend our research into communication resource allocation, addressing the challenges of spectral efficiency, interference management, and hybrid beamforming optimization in C-NOMA-enabled multi-UAV mmWave networks. We introduce a joint optimization framework that integrates hybrid beamforming (HBF) and power allocation (PA) strategies, ensuring efficient subchannel reuse while minimizing inter-cluster interference. The problem is formulated as a multi-agent reinforcement learning (MARL) task, leveraging a deep deterministic policy gradient (MADDPG) algorithm to optimize beamforming and power control dynamically. Additionally, we explore multi-agent collaboration for adaptive hybrid beamforming, where UAVs coordinate power allocation, beam directions, and subchannel assignments in real time, significantly enhancing energy and spectral efficiency.
This thesis contributes to the advancement of 6G-enabled IIoT networks by presenting innovative solutions for hierarchical task offloading, energy-efficient MEC service provisioning, and communication resource optimization in multi-UAV-assisted air-ground networks. The proposed architectures, optimization frameworks, and AI-driven decision-making algorithms provide a scalable and efficient foundation for next-generation 6G networks.
| Date of Award | 2025 |
|---|
| Original language | American English |
|---|
| Awarding Institution | - HBKU College of Science and Engineering
|
|---|